Enhancing a Pairs Trading strategy with the application of Machine Learning

作者：

Highlights：

• The application of Unsupervised Learning is advantageous to find promising pairs.

• A proposed forecasting-based trading model reduces the period of portfolio decline.

• Profitability is affected when long formation periods are required.

• Trading commodity-linked ETFs in a 5-min setup proves auspicious.

• Selecting pairs based on validation performance is beneficial.

摘要

•The application of Unsupervised Learning is advantageous to find promising pairs.•A proposed forecasting-based trading model reduces the period of portfolio decline.•Profitability is affected when long formation periods are required.•Trading commodity-linked ETFs in a 5-min setup proves auspicious.•Selecting pairs based on validation performance is beneficial.

论文关键词：Pairs trading,Market neutral,Machine Learning,Deep learning,Unsupervised learning

论文评审过程：Received 6 November 2019, Revised 10 March 2020, Accepted 27 April 2020, Available online 4 May 2020, Version of Record 20 May 2020.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.113490